+1 vote
asked by (170 points)

I'm attempting to multiply two MPOs. I do, in full,

#include "itensor/itensor.h"
#include "itensor/mps/autompo.h"
#include "itensor/mps/sites/spinhalf.h"
#include "itensor/util/print_macro.h"

using namespace itensor;

int main(int argc, char* argv[])
  int L      = 8;
  float hz   = 2.0;

  auto sites = SpinHalf(L);
  auto H_ampo = AutoMPO(sites);

  for(int j = 1; j <=  L; j++) { H_ampo += hz, "Sz",j; }

  MPOt<ITensor> H  = MPO(H_ampo);
  MPOt<ITensor> Hn = MPO(H_ampo);
  MPOt<ITensor> res = MPO(H_ampo);


  nmultMPO<MPOt<ITensor>>(H, Hn, res);


I compile with a Makefile only slightly modified from the sample---I'd be happy to post it, but it doesn't seem terribly interesting.

After printing something reasonable for the indices of H.A(2), the thing allocates hundreds of mb of memory (per top) and then runs for a long time (tens of minutes) before returning. This is not what I expect for an eight-site, bond-dimension-1 MPO!

I'm sure I'm misunderstanding something, here. What is it?

1 Answer

0 votes
answered by (70.1k points)

Hi, sorry about the slow reply! This is a good question: so nmultMPO expects additional named arguments saying what truncation parameters to use. Without them it was keeping exponentially many states as it went from left to right "compressing" the two MPOs.

I just pushed a fix for this in the latest version of ITensor. Now if you don't state any truncation parameters it defaults to a small cutoff of 1E-14.

But the best way to call this algorithm is as:


where you may want to use a larger or smaller cutoff than 1E-12; it depends on the operators and what you need them for.

By the way, I would strongly recommend against using the float type in C++; use double instead or Real which is defined in ITensor to be double.

Also you don't need to write MPOt. The type MPO is an alias for this. Even better you can write auto H = MPO(H_ampo); and the compiler will deduce that the object on the left has the same type as on the right. Similarly you don't need to provide template parameters to call nmultMPO; the compiler will deduce the template parameters of functions for you whenever they are unambiguous.

Please let me know if you have additional questions -


commented by (70.1k points)
Here is an updated version of your code with the recommended C++ style:

int main()
    int  L   = 8;
    Real hz  = 2.0;
    Real cutoff = 1E-12;

    auto sites = SpinHalf(L);
    auto H_ampo = AutoMPO(sites);

    for(int j = 1; j <=  L; j++)
        H_ampo += hz, "Sz",j;

    auto H  = MPO(H_ampo);
    auto Hn = MPO(H_ampo);
    auto res = MPO{};


    nmultMPO(H, Hn, res,{"Cutoff",cutoff});

    return 0;
Welcome to ITensor Support Q&A, where you can ask questions and receive answers from other members of the community.

Formatting Tips:
  • To format code, indent by four spaces
  • To format inline LaTeX, surround it by @@ on both sides
  • To format LaTeX on its own line, surround it by $$ above and below
  • For LaTeX, it may be necessary to backslash-escape underscore characters to obtain proper formatting. So for example writing \sum\_i to represent a sum over i.
If you cannot register due to firewall issues (e.g. you cannot see the capcha box) please email Miles Stoudenmire to ask for an account.

To report ITensor bugs, please use the issue tracker.